9
\$\begingroup\$

I have a question regarding dc-dc buck converters. I need to supply multiple different voltages on my board and I am very limitted with power this is why I need to use high efficiency converters. If they are to be placed near to each other should their switching frequency be the same? As far as I know this would be better due to the EMI and SI issues, it's easier to get rid off one frequency than multiple of them and their harmonics, products etc.

On the other hand if it's not essential, than increasing the switching frequency will decrease the size of the inductors.

Any help would be appreciated

\$\endgroup\$
3
  • \$\begingroup\$ FYI. Not all of the buck converters/controllers have a fixed frequency. Many have a varying frequency, which varies with load. What buck converter model do you have in mind? \$\endgroup\$ Apr 13 '13 at 1:31
  • \$\begingroup\$ Couple more questions. Do you have RF communication components on your board? What kind of EMI compliance do you have to do? \$\endgroup\$ Apr 13 '13 at 1:42
  • \$\begingroup\$ I have TPS40041 from TI in mind or/and XRP7724. In the first one the switching frequency is fixed at 600KHz in the other IC I can adjust it. For clarity: Multiple dc-dc converters will have separate power supply rails. I should've added that information, sorry \$\endgroup\$
    – zdun8
    Apr 13 '13 at 14:52
6
\$\begingroup\$

As far as I know this would be better due to the EMI and SI issues, it's easier to get rid off one frequency than multiple of them and their harmonics, products etc.

This is a false premise. EMI regulations limit the emissions on a frequency-by-frequency basis. If you have two sources in your system at the same frequency, their outputs can add, giving higher emissions at that frequency. If they're at different frequencies, they are effectively independent for purposes of emissions.

It's a general rule in EMI design that the best solution is to reduce your sources of emissions, rather than generate signals and then try to block them. So I'd say for EMI purposes, you're better off using different frequencies for your different switching regulators.

\$\endgroup\$
3
  • 1
    \$\begingroup\$ There is even a technique to spread out the spectrum by varying switching frequency. This way it is easier to pass EMI certification. \$\endgroup\$ Apr 13 '13 at 10:28
  • \$\begingroup\$ I see, I thought it works the other way round. I will set the frequencies differently in the converters. Thank you. \$\endgroup\$
    – zdun8
    Apr 13 '13 at 15:00
  • \$\begingroup\$ @zdun8, as Szymon points out in his answer, if your regulators allow synchronization (with different phases) you can actually use this to make the two regulators interfere destructively and reduce emissions. But if your regulators don't have a sync feature, you're better off using different frequencies. \$\endgroup\$
    – The Photon
    Apr 13 '13 at 15:05
3
\$\begingroup\$

If you use multiple converters with different switching frequencies it will be really hard to predict the input voltage ripple and thus it will be hard to design the input filter. There will be some moments when switching occurs simultaneously and some moment when switching events will be spread out in time.

In your case, I think, the best design would be to use the same switching frequency for all converters and interleave them. This way the input filter for all buck will be much smaller than the sum of individual filters (if they would not be interleaved).

\$\endgroup\$
1
\$\begingroup\$

Using independent frequencies probably has no massive negative aspect.

Identical frequencies, if they really are identical, may lead to switching transients from one converter getting into another at a critical part in the switching cycle and affecting when and how it switches. At the likely levels involved in cross feeding I would expect this usually not to be a fatal problem - just MAY degrade accuracy of output if switching point is slightly affected by other signals.

Such spurious input would usually have effect only where they affected the switching decision voltage when it was very near to the switching threshold as at other points in the cycle the voltages will be large enough that noise will not affect them. eg if switching point occurs when divided down output is fed to Vref pin and = Vref = say 0.8V, then if Vin is 0.799 V, noise on the sense line that is coupled onto Vsense and that results in a + 0.001V change may make it switch early. But if Vsense is at 0.700 Volts you'd need +0.1V noise to trigger switching

Higher frequency Noise on Vout has a fairly free ride on to Vsense as there is usually a cap across the reference divider from Vsense to Vout. This vastly improves response time to transients and removal can lead to a converter which is fatally trying to chase its own tail. Ask me how I know :-).

If N converters are asynchronous wrt frequency etc then glitches will occur pseudo randomly throughout the cycle and hopefully have little overall effect.

\$\endgroup\$
1
\$\begingroup\$

If you run your bucks at different frequencies, you need to be careful of beat frequencies - beating shows up as harmonic content on the input DC, and if unfiltered, the buck loop compensation may not handle it and allow it to pass to the output without attenuation.

Good input filtering is important (properly located high frequency ceramic and/or film capacitors near the input of each converter) as well as good layout practices (keeping the switching nodes as tight as possible, keeping physical separation between the bucks and not mixing the power/control routes between them, etc).

\$\endgroup\$
1
\$\begingroup\$

My experience: A particular job that had two switchers locked to the same frequency and then, on a 2nd iteration PCB mistakingly leaving a component out that "locked" them made no difference to the baseband noise whatsoever; neither did it affect power efficiency.

The job was quite analoggy with several strain gauge amplifiers etc..

As far as EMC goes, like @The Photon says, use two frequencies because emissions are likely to be less, and this is supported by the manufacturers of several SMPSU chips who modulate their switching frequency to avoid this.

\$\endgroup\$

Not the answer you're looking for? Browse other questions tagged or ask your own question.